Essays in Empirical Operations Management

Other Titles

This dissertation focuses on the empirical studies in two streams of operations management - supply chains and marketplaces. The first and third chapters are joint work with Professor Vishal Gaur. The second chapter is joint with Prof. Fanyin Zheng. In chapter 1, we use firm-level financial data from S&P's Compustat to study the problem of managing and forecasting a firm's cash flow from operations. The study aims to equip a financial manager or a C.F.O. with a range of operational policies that they may evaluate to improve their firm's cash flow. This problem, however, turns out to be notoriously hard for the following reason. Cash flow comprises multiple operational variables (sales, inventory, etc.) that are endogenously determined and interact dynamically. As such, it is hard to cleanly estimate the impact of any prescriptions for managing cash flows. To solve this, we built a joint data-driven structural model of the endogenous evolution of these variables. We characterize the system's short- and long-run responses to shocks in different operational variables using these estimates. The research output has clear, practical implications. Managers can directly feed these responses into their firm's decision support toolbox as input. In addition, they can use our model to forecast cash flows, which we show outperforms univariate models in its forecasting ability. Finally, the model also allows a manager to estimate the impacts of periods of economic recession on their firm's financial performance in the future. In chapter 2, we provide price-policy- and operational-policy-based prescriptions to a city government that wishes to reduce congestion and greenhouse gas emissions. To do so, we first model customers' decisions of choosing on-demand vs. scheduled shuttle rides. We combined multiple large-scale and micro-level datasets in this work, for example, a company-provided dataset with over 25 million customer ride details; and a google API-based dataset containing hyperlocal information about 500k points of interest in the city. Using these data, we estimate how different customers value inconveniences associated with pooled rides across various times in a day. We use these estimates to evaluate the policies mentioned above for reducing congestion. We find that costless operational improvement policies in the shuttle service design can achieve a substantial amount of traffic congestion reduction without harming customer welfare. Chapter 3 investigates the relative efficacy of data- and model-parameter-based clustering methods for forecasting key economic quantities of interest in an operations setting, like sales, ridership, and inventories. In a panel data setting, we show that model-parameter based clustering outperforms data-based clustering method in its forecasting ability. Moreover, model-parameter based clustering also outperforms pooled- and individual unit level models. We demonstrate our results via a simulation on panel data and two empirical applications on retail sales forecasting and NYC citibike ridership. We finally provide theoretical and econometric arguments to support our analysis.

Journal / Series
Volume & Issue
151 pages
Date Issued
Effective Date
Expiration Date
Union Local
Number of Workers
Committee Chair
Gaur, Vishal
Committee Co-Chair
Committee Member
Henderson, Shane G.
Gavirneni, Nagesh
Degree Discipline
Degree Name
Ph. D., Management
Degree Level
Doctor of Philosophy
Related Version
Related DOI
Related To
Related Part
Based on Related Item
Has Other Format(s)
Part of Related Item
Related To
Related Publication(s)
Link(s) to Related Publication(s)
Link(s) to Reference(s)
Previously Published As
Government Document
Other Identifiers
Rights URI
dissertation or thesis
Accessibility Feature
Accessibility Hazard
Accessibility Summary
Link(s) to Catalog Record